Computer vision-based analysis of images in university-curated Instagram posts: a user-engagement perspective

Abstract

Higher education institutions increasingly depend on images to increase user engagement in their social media posts. The aim of this study is hence to help identify image variables and characteristics to offer strategic insights. This study analysed 342 images embedded in high-engagement Instagram posts of 25 top-ranked universities using the Google Cloud Vision object detection feature. The examination yielded a collective occurrence frequency of 3234 objects, detecting 104 different labels. ‘Person’ followed by ‘clothing’ were identified to be the foremost object labels in the sampled images. A subsequent qualitative classification of the more prominent image objects revealed four most common types of images: personal attire (53.9%), social presence (34.5%), university servicescapes (10.4%), and endearing pets (1.1%). These themes are discussed with a consumer value conceptual lens for their depiction of utilitarian, emotional, social, symbolic, exclusivity and experiential value types. The present study is an opening effort to incorporate a computer vision-based methodology to analyse images in university-curated social media posts for consumer engagement perspectives.